A constrained-average strategy for reduction of artifacts from scattering inhomogeneities in parametric images of the attenuation coefficient

This work investigates a strategy to reduce artifacts caused by macroscopic scattering discontinuities when forming parametric images of the ultrasound attenuation coefficient of soft tissue. If present, these artifacts can cause mis-correlation with the envelope image and introduce errors in subsequent assessments of tissue backscatter coefficients. Results using the proposed strategy are compared to conventional attenuation estimations on images of tissue mimicking phantoms with scattering inhomogeneities. Four phantoms containing targets with circular cross sections and different scattering properties from the surrounding material (background) were scanned. Five uncorrelated RF data frames were acquired from each target and from the background with a Siemens S2000 system. Attenuation images were formed by applying the Reference Phantom/Spectral Difference Method within estimation blocks shifted across the data frame. Conventionally, each attenuation estimate is mapped to the location of the center of the estimation block. In the proposed “constrained-average” strategy, the estimates from all blocks that include a given power spectral density window are averaged, and the result is assigned to the window location, only considering estimates within the range 0 <; α₀ <; 4 dB/cm-MHz, where α₀ is the slope of the attenuation coefficient vs. frequency. Results show that conventional attenuation images are more sensitive to scattering discontinuities than the constrained strategy, leading to values biased by up to 40%. The constrained average strategy reduced the bias and the variability of the α₀ values within the structures by about 20%.